Swiss researchers have created teams of "antbots", electronic robots that interact and co-operate in an ant-like way, sharing resources and helping one another to forage for food. The work is described in a recent issue of Nature, which also carries a report about robots that design and build themselves without human intervention.
"Efficient mechanisms of division of labour, in particular series-parallel operation and transfer of information among group members, are key components of the tremendous ecological success of ants," the authors write.
The creation of Dr Laurent Keller and colleagues at the University of Lausanne, the antbots were taught community spirit by being programmed with just a few of the rules that ants use to co-ordinate their foraging. These include when to leave the nest, how to avoid crashing into one another and, most importantly, how to tell their nest mates if they need help retrieving food clusters.
The antbots were set on a large flat surface and their ability to forage was measured as an indication of their success as a "species". Sure enough, the antbots working in teams were better able to find and retrieve "food" than individuals working on their own. But the researchers found the benefits of teamworking declined when the group was too large. "Intriguingly, a similar relationship between group size and efficiency has been documented in social insects," the authors said.
Antbots with the ability to recruit help from others after discovering a large cache of food were also found to be more effective foragers than those that couldn't communicate.
The findings, the researchers said, have important implications in robotics, particularly where robots must perform effectively in dangerous places or where the risk of system failure must be avoided - for example, during robotic missions to Mars.
They picture a situation where robot teams could be placed on a distant planet, and would collect information and support one another while out of the reach of human operators.
Automated design and manufacture of robotic life forms was the goal of a research team from Brandeis University in Massachusetts. They created a computer programme that would allow robot forms to evolve spontaneously, leading to robots that could move about by themselves.
The team specified simple building blocks for robot life forms, including bars and actuators as building blocks of structure, and "artificial neurons" as building blocks of control. The bars in turn were connected by ball and socket joints that could form rigid or articulating joints. Success was measured by the locomotive ability of the robot design, a sign of relative "fitness". The researchers were determined that the systems should remain autonomous to prevent human operators from overcoming difficulties or taking shortcuts on behalf of the computer.
Once the building blocks were defined, an initial population of 200 such machines with "zero" bars and "zero" neurons were allowed to evolve through between 300 and 600 generations.
This allowed the body and the brain of the robots to co-evolve. The computer selected the fittest robots for further evolution. These were then automatically taken out of their virtual world and converted into physical objects.
Finally, human operators snapped in actuators as specified by the computer, and three working, moving robots were let loose on the world. The researchers being particularly struck by their remarkable symmetry, "which was neither specified nor rewarded for anywhere in the code".
Their work is praised in an associated News and Views report by Dr Rodney Brooks of MIT, who said they had bridged the gap between computer models and physical reality. "This is the long awaited and necessary step towards the ultimate dream of self-reproducing machines," he said.